Fix some errors from go code analyzers in docker-plugin

…ant box.

root@noiro-ucs08:~/cilium-net/packer-scripts/ubuntu-14.04# ls -l *.box
-rw-r--r-- 1 root root 364980917 Dec 29 15:31 ubuntu-1404-libvirt.box

vagrant@ubuntu:~$ uname -a
Linux ubuntu 4.1.0-rc4-custom #1 SMP Tue Dec 29 15:26:42 PST 2015 x86_64 x86_64 x86_64 GNU/Linux
vagrant@ubuntu:~$

Fix some errors from go code analyzers in docker-plugin

When the allocation CIDR changes, the existing code errored out. When that
happened on the first agent restart, all existing endpoints failed to restore
due to failure to allocate the existing endpoint's IP address. The failure to
restore would then in turn get all existing endpoints deleted.

This is bad in two ways:

1. It causes all endpoints to be removed, causing the pods to restart
2. It does not allow the agent to ever come up again without intervention from
   the user.

This commit solves problem #2 to allow nodes to recover automatically. It does
not solve problem #1 yet and requires pods to restart.

Solving problem #2 only makes sense once we have isolated all datapath aspects
such as the DNS proxy and L7 proxies out of the agent as endpoint connetivity
cannot be guaranteed if the agent cannot come up within a reasonable amount of
time.

Fixes: #6215

Signed-off-by: Thomas Graf <thomas@cilium.io>

When the allocation CIDR changes, the existing code errored out. When that
happened on the first agent restart, all existing endpoints failed to restore
due to failure to allocate the existing endpoint's IP address. The failure to
restore would then in turn get all existing endpoints deleted.

This is bad in two ways:

1. It causes all endpoints to be removed, causing the pods to restart
2. It does not allow the agent to ever come up again without intervention from
   the user.

This commit solves problem #2 to allow nodes to recover automatically. It does
not solve problem #1 yet and requires pods to restart.

Solving problem #2 only makes sense once we have isolated all datapath aspects
such as the DNS proxy and L7 proxies out of the agent as endpoint connetivity
cannot be guaranteed if the agent cannot come up within a reasonable amount of
time.

Fixes: #6215

Signed-off-by: Thomas Graf <thomas@cilium.io>

[ upstream commit 0b3ed95 ]

When the allocation CIDR changes, the existing code errored out. When that
happened on the first agent restart, all existing endpoints failed to restore
due to failure to allocate the existing endpoint's IP address. The failure to
restore would then in turn get all existing endpoints deleted.

This is bad in two ways:

1. It causes all endpoints to be removed, causing the pods to restart
2. It does not allow the agent to ever come up again without intervention from
   the user.

This commit solves problem #2 to allow nodes to recover automatically. It does
not solve problem #1 yet and requires pods to restart.

Solving problem #2 only makes sense once we have isolated all datapath aspects
such as the DNS proxy and L7 proxies out of the agent as endpoint connetivity
cannot be guaranteed if the agent cannot come up within a reasonable amount of
time.

Fixes: #6215

Signed-off-by: Thomas Graf <thomas@cilium.io>
Signed-off-by: Ian Vernon <ian@cilium.io>

[ upstream commit 0b3ed95 ]

When the allocation CIDR changes, the existing code errored out. When that
happened on the first agent restart, all existing endpoints failed to restore
due to failure to allocate the existing endpoint's IP address. The failure to
restore would then in turn get all existing endpoints deleted.

This is bad in two ways:

1. It causes all endpoints to be removed, causing the pods to restart
2. It does not allow the agent to ever come up again without intervention from
   the user.

This commit solves problem #2 to allow nodes to recover automatically. It does
not solve problem #1 yet and requires pods to restart.

Solving problem #2 only makes sense once we have isolated all datapath aspects
such as the DNS proxy and L7 proxies out of the agent as endpoint connetivity
cannot be guaranteed if the agent cannot come up within a reasonable amount of
time.

Fixes: #6215

Signed-off-by: Thomas Graf <thomas@cilium.io>
Signed-off-by: Ian Vernon <ian@cilium.io>

…mplementation-in-cilium

Add Gateway API datapath walkthrough

fix: set illustration absolute in Hero

feat: add Blog search

* re-write (cilium#1)

* wording